In [40]:
from collections import Counter
## checking data structure and class distribution
# Count the number of samples per class in each dataset
def count_classes(dataset):
class_counts = Counter()
for _, labels in dataset:
class_counts.update(labels.numpy().tolist())
return class_counts
# Analyze class distribution
train_class_counts = count_classes(train_dataset)
val_class_counts = count_classes(validation_dataset)
test_class_counts = count_classes(test_dataset)
print("Class distribution in training set:", train_class_counts)
print("Class distribution in validation set:", val_class_counts)
print("Class distribution in test set:", test_class_counts)
# Visualize class distribution
import matplotlib.pyplot as plt
labels = ['Cat', 'Dog']
train_counts = [train_class_counts[0], train_class_counts[1]]
val_counts = [val_class_counts[0], val_class_counts[1]]
test_counts = [test_class_counts[0], test_class_counts[1]]
plt.figure(figsize=(12, 6))
plt.bar(labels, train_counts, label='Train', alpha=0.6)
plt.bar(labels, val_counts, label='Validation', alpha=0.6)
plt.bar(labels, test_counts, label='Test', alpha=0.6)
plt.title("Class Distribution Across Train, Validation, and Test Sets")
plt.ylabel("Number of Samples")
plt.legend()
plt.show()
Class distribution in training set: Counter({1: 1000, 0: 1000})
Class distribution in validation set: Counter({0: 500, 1: 500})
Class distribution in test set: Counter({1: 1000, 0: 1000})
